Software Beta Testing Question

[Jeff]

I use a program called Spectrum SentryNet to control all of my scanners here at the house. SentryNet

The folks that own the company are asking me to do some Beta Testing for them with a new add on pack for this software. The add on incorporates a "Speech Prediction Algorithm". It will use a speech estimator to search frequency ranges and discriminate between human voice and noise or data signals.

I have not started doing this yet and I have to wait until I get the add-on pack sent here. In the meantime I have been trying to do a little research on the net and see if I could find any websites that might give me some basic familiarity with the concepts behind this.

I have not had allot of luck finding too much. I figured I would throw out a post and see if anyone on here might have some knowledge of the concept and might know of a few links they could throw at me off the top of their head that might help me brush up on the subject.

Any websites the deal with this would be most appreciated. In the meantime I'll be doing more looking myself.

Thanks a a heap.

 

[The_Twisted_Ear]

CELP means Code Excited Linear Prediction. Cell phones use speech prediction algorithm to digitize the voice, but this algorithm works poorly on anything but human speech.

Jump over to Google and enter (with quotes): "Code Excited Linear Prediction"

 

[Jeff]

CELP means Code Excited Linear Prediction. Cell phones use speech prediction algorithm to digitize the voice, but this algorithm works poorly on anything but human speech.

Jump over to Google and enter (with quotes): "Code Excited Linear Prediction"

Thanks. I have been playing with the software and it is so buggy I have not had a chance to mess with this Speech Prediction Estimator as yet. I need the software to work half way decently doing what I am used to seeing it do before I try to get it to take on these advanced featurs.

 

[RadioCtrlDWife]

Requantitization

Thanks. I have been playing with the software and it is so buggy I have not had a chance to mess with this Speech Prediction Estimator as yet. I need the software to work half way decently doing what I am used to seeing it do before I try to get it to take on these advanced featurs.

Hi Jeff
One issue I foresee it is, that because this encoding algorithm was designed [I think by Qualcom] to compress speech data on the fly from a microphone into low bit rate data; it may not work well on "second pass digitization' of the scanner's analog demodulated output and requisite requanitization by a sound card in the PC to digitize for processing. Then again, you only need to see if the audio is speech or noise, so you may have no issues.
As always, Jeff I am ready to help if needed. You provide a valuable service in the absence of other instant local information about public safety.

 

[Jeff]

Hi Jeff
One issue I foresee it is, that because this encoding algorithm was designed [I think by Qualcom] to compress speech data on the fly from a microphone into low bit rate data; it may not work well on "second pass digitization' of the scanner's analog demodulated output and requisite requanitization by a sound card in the PC to digitize for processing. Then again, you only need to see if the audio is speech or noise, so you may have no issues.
As always, Jeff I am ready to help if needed. You provide a valuable service in the absence of other instant local information about public safety.

I have not gotten into this new stuff as yet as the basic software does appears to have it's issues. I need to have it working correctly with some margin of confidence before trying to get into the nuts & bots of the newly added features. Thus far I the new changes made to the software seem to have made a significant impact of the way everything else runs. Numerous crashes, it records audio on frequencies when I have not told it to, and a few other minor quirks.

I think the guy is getting a bit impatient with me because he wants me to test this new feature out for him so he can sell it to the Govt. But I am not even touching the new feature until the basic software will run with some degree of certainty.

I have shot a few emails back & forth with them and they way they explain the new feture to me is that I am supposed to be able to SEARCH a given frequency range with this feature enabled. I have to be honest in that the feature itself is a bit above my head but once I get into it, it should all fall into place a bit better. At this point I am not even exactly sure what it is suppposed to do.

 

[RadioCtrlDWife]

Peak Limit the Audio

I have not gotten into this new stuff as yet as the basic software does appears to have it's issues. I need to have it working correctly with some margin of confidence before trying to get into the nuts & bots of the newly added features. Thus far I the new changes made to the software seem to have made a significant impact of the way everything else runs. Numerous crashes, it records audio on frequencies when I have not told it to, and a few other minor quirks.

I think the guy is getting a bit impatient with me because he wants me to test this new feature out for him so he can sell it to the Govt. But I am not even touching the new feature until the basic software will run with some degree of certainty.

I have shot a few emails back & forth with them and they way they explain the new feture to me is that I am supposed to be able to SEARCH a given frequency range with this feature enabled. I have to be honest in that the feature itself is a bit above my head but once I get into it, it should all fall into place a bit better. At this point I am not even exactly sure what it is suppposed to do.

Jeff, I have an idea for you to try. I will lend you an old, but low distortion, analog studio peak limiter [actually a classic UREI BL-40] to remove some of the crest factor of the static and to keep the audio within the optimum dynamic range window of the sound card input. This will help the quantization process by keeping the ENOB [effective number of bits] high with varying audio levels, reducing quantization errors to a minimum.
Here is an "instrument" to help you set up the interface
Personal Embedded RTA.

Have a Five Bar Linear Evening!

 

[Jeff]

Jeff, I have an idea for you to try. I will lend you an old, but low distortion, analog studio peak limiter [actually a classic UREI BL-40] to remove some of the crest factor of the static and to keep the audio within the optimum dynamic range window of the sound card input. This will help the quantization process by keeping the ENOB [effective number of bits] high with varying audio levels, reducing quantization errors to a minimum.
Here is an "instrument" to help you set up the interface
Personal Embedded RTA.

Have a Five Bar Linear Evening!

As I am Beta Testing his stuff for him at this point I would not want to bring any additional stuff into the mix to further complicate things. There are enough bugs in the software as is without adding another doo-dad to the mix.

That shirt looks a bit complicated for me. Cool as heck though.